At air cargo terminals, the operations of an automatic storage and retrieval system (AS/RS) have certain special characteristics that both cargo arrival rates and storage duration are stochastic, and the probability distributions can only come from the empirical data. According to such particularities, this paper proposes a class-based reservation storage policy for AS/RS at air cargo terminals and an analytical model is developed. Two classes are investigated and a certain storage spaces are reserved for one priority class in such a way that the one-way travel time of S/R machine is minimized. The optimal reservation space is obtained by numerical searching. A simulation model is developed and the simulation results validate the optimal solution from the analytical model.